PASCAL - Pattern Analysis, Statistical Modelling and Computational Learning

Blind Identification of Overcomplete Mixture of Sources (BIOME)
Laurent Albera, Anne Ferreol, Pierre Comon and Pascal Chevalier
Linear Algebra and Applications Volume 391, pp. 1-30, 2004. ISSN 0024-3795

Abstract

The problem of Blind Identification of linear mixtures of independent random processes is known to be related to the diagonalization of some tensors. This problem is posed here in terms of a non-conventional joint approximate diagonalization of several matrices. In fact, a congruent transform is applied to each of these matrices, the left transform being rectangular full rank, and the right one being unitary. The application in antenna signal processing is described, and suboptimal numerical algorithms are proposed. Keywords: Blind Source Separation; Independent Component Analysis; Tensor; High Order Statistics; Joint Diagonalization; Underdetermined linear system; Overcomplete representation

EPrint Type:Article
Project Keyword:Project Keyword UNSPECIFIED
Subjects:Computational, Information-Theoretic Learning with Statistics
Learning/Statistics & Optimisation
Theory & Algorithms
ID Code:1641
Deposited By:Pierre Comon
Deposited On:28 November 2005